N-Terminal or C-Terminal? Where to Put Your Tag (and Why It Changes Everything)

You designed two constructs of the same protein with the same His10 tag. The only difference: one carries the tag at the N-terminus, the other at the C-terminus. The N-terminal version expresses at 40 mg/L and purifies clean. The C-terminal version gives you a faint smear and a protein that has quietly lost the activity you needed to measure. Same gene, same tag, same host, same day at the bench. Opposite outcome. The variable you changed was not the tag. It was the end you put it on.

Terminus placement is the quietest decision in construct design and one of the most consequential. It sits upstream of expression level, folding, secretion, membrane insertion, function, and how cleanly you can purify. Most people default to the N-terminus because that is where the vector's tag happens to sit, and then spend three weeks at the bench discovering why that was wrong for this particular protein.

Key Takeaways

• The terminus is a separate decision from the tag. Which handle you purify by (His vs Strep vs FLAG) is one choice; which end you attach it to is a different one that affects expression, folding, and function independently.
• The N-terminus is where expression is decided. In E. coli, mRNA secondary structure and codon choice in the first ~30 codons drive expression more than the tag itself, changing yield by up to ~14-fold in controlled measurements (Goodman et al., 2013).
• Your N-terminal residue sets protein stability. The N-end rule links the identity of the N-terminal amino acid to intracellular half-life, from minutes to more than twenty hours (Bachmair et al., 1986).
• Some placements are non-negotiable. Signal peptides and most solubility fusions must be N-terminal; CAAX prenylation, PDZ-binding tails, KDEL ER-retention, and tail-anchor membrane targeting all live at the C-terminus and a C-terminal tag destroys them.
• Membrane topology decides accessibility. Whether a terminus is cytoplasmic or periplasmic determines whether a tag folds, gets captured, or flips the protein's topology (Daley et al., 2005).
• A C-terminal tag reports on full-length protein only. Because translation runs N to C, capturing by a C-terminal tag enriches for molecules that were made all the way through.

Terminus Is Not the Same Decision as Tag

First, scope, because two nearby decisions get conflated. Choosing which affinity handle goes on your protein (His6 vs Strep-tag II vs FLAG, and their tradeoffs in selectivity, elution, and cost) is a separate article, and we covered it in Choosing an Affinity Tag. This post assumes you have already picked the tag. The question here is where it goes: N-terminus or C-terminus.

That placement is not cosmetic. The two termini of a protein are chemically and biologically different environments. The N-terminus is translated first, carries the initiator methionine, sets the protein's degradation signal, and is the mandatory location for signal peptides. The C-terminus is translated last, is the reporter of complete synthesis, and is where a large fraction of eukaryotic targeting and interaction motifs live. Put a tag on the wrong one and you are not adding an inert handle, you are overwriting biology.

The good news: both termini are, on average, solvent-exposed. A survey of PDB structures found N- and C-termini sit predominantly on the protein surface, which is why tagging works at all (Jacob & Unger, 2007). The bad news: "on average" is doing a lot of work. For your specific protein, one terminus may be buried, functional, or load-bearing, and the whole game is knowing which.

Expression Is Decided at the N-Terminus

In E. coli, the single largest determinant of whether a gene expresses is not the promoter, the codon optimization of the whole ORF, or the tag. It is what happens in roughly the first 30 codons.

The mechanism is mRNA secondary structure around the ribosome binding site and start codon. Stable hairpins in the 5' region occlude the Shine-Dalgarno sequence and the start codon, and the ribosome cannot initiate efficiently on a folded message. Kudla et al. (2009) synthesized 154 GFP variants that were synonymous (identical protein, different codons) and found that expression varied over a 250-fold range, correlating with the predicted folding energy of the 5' mRNA, not with codon bias across the whole gene.

Goodman et al. (2013) sharpened this with more than 14,000 synthetic reporters and showed that using rarer codons in the N-terminal region actually raised expression, by up to ~14-fold, precisely because rare codons tend to reduce local mRNA structure. The lesson is not "use rare codons." The lesson is that the N-terminal region is where translation initiation is won or lost.

Why this matters for tag placement: an N-terminal tag inserts a fixed peptide sequence directly into this critical zone. That sequence, and the DNA encoding it, becomes the new 5' context of your message. A poorly chosen N-terminal tag or linker can introduce a hairpin that shuts down initiation, while a C-terminal tag sits far downstream and leaves the initiation region untouched. This is a genuine argument for C-terminal placement when expression level is the thing you are fighting for, and it is an argument for codon-optimizing the tag-plus-linker region deliberately rather than pasting in a stock sequence.

The N-End Rule: Your N-Terminal Residue Sets Half-Life

The N-terminal residue of a protein is a degradation signal. The N-end rule, established when Bachmair, Finley, and Varshavsky (1986) engineered otherwise-identical beta-galactosidase reporters differing only in their N-terminal residue, links that residue to intracellular half-life. Some N-terminal residues gave a half-life of more than 20 hours; others reduced it to under 3 minutes.

The pathway is now understood in detail across bacteria and eukaryotes (Varshavsky, 2011): "stabilizing" residues (Met, and in bacteria Ala, Ser, Thr, Val, and others) leave the protein alone, while "destabilizing" residues (Arg, Lys, Leu, Phe, and more) recruit the ubiquitin or ClpAP machinery.

This interacts with tag placement two ways. First, an N-terminal tag defines the N-terminal residue: most tags begin with Met after the initiator, which is stabilizing, so an intact N-terminal tag usually protects the protein. Second, and more subtly, the residue you expose after cleaving an N-terminal tag is set by your protease. TEV leaves a serine or glycine (stabilizing); some strategies expose a destabilizing residue. If you plan to remove an N-terminal tag, check what residue you will leave behind, because you may be converting a stable protein into a rapidly degraded one.

One more N-terminal wrinkle: methionine aminopeptidase (MAP) removes the initiator Met when the second residue is small (Ala, Cys, Gly, Pro, Ser, Thr, Val). Frottin et al. (2006) mapped this specificity across the proteome. If your N-terminal tag design assumes the Met stays (or assumes it goes), the actual processed N-terminus can differ from what you drew, which matters for mass spec, for the N-end rule, and for anything that needs a defined N-terminus.

Solubility: Does the Terminus Change It?

Yes, and not always in the direction you would hope. In a controlled study across many human targets, Woestenenk et al. (2004) found that both N-terminal and C-terminal His6 tags had a noticeable negative effect on solubility, and critically, the effect was target-protein-specific: some proteins tolerated one terminus and not the other, with no universal winner.

The practical reading: a small affinity tag does not reliably rescue solubility at either end, and it can hurt. If solubility is your problem, you are no longer choosing a terminus for a small tag, you are choosing where to put a solubility fusion, and that decision is heavily biased.

Solubility fusions like MBP, SUMO, NusA, and Trx work partly by folding first and nucleating the folding of the downstream passenger. That mechanism only operates if the fusion is translated first, which means N-terminal placement is strongly preferred for solubility fusions. A C-terminal MBP cannot chaperone a passenger that has already folded (or misfolded) before MBP appears. This is why nearly every validated solubility-fusion construct is N-terminal, a point we develop in the fusion-partner guide. If your protein needs a folding chaperone fused to it, the terminus is effectively chosen for you.

Folding Interference: Which Terminus Is Safe to Load

Even an inert tag adds mass, charge, and a linker at one end of the fold. Whether that perturbs folding depends on what that terminus is doing structurally. Ask three questions before you tag it:

1. Is it buried? Termini are usually exposed, but not always. A buried terminus tolerates a tag poorly and may not present it for capture.
2. Is it near the folding core or a domain interface? A tag adjacent to a structured element can slow or block folding, especially during co-translational folding where the nascent chain folds as it emerges.
3. Is it close to the active site or a binding surface? Proximity means the tag can occlude function without ever touching the catalytic residues.

The N-terminus emerges from the ribosome first and folds co-translationally, so a bulky N-terminal fusion can interfere with early folding (or stabilize an exposed nascent chain). The C-terminus folds last and is often more forgiving of added mass, which is one reason C-terminal reporters are so common. There is no universal answer, which is exactly why predicting disorder and structural context at each terminus before cloning beats finding out on a gel.

Signal Peptides and Secretion: N-Terminal Only

If you want your protein secreted or exported to the periplasm, a signal peptide directs it, and signal peptides are strictly N-terminal. They are short, N-terminal sequences with a positively charged n-region, a hydrophobic h-region, and a cleavage site recognized by signal peptidase, a tripartite architecture established across the classic signal-sequence literature. The signal recognition and translocation machinery reads the N-terminus of the emerging chain, so the signal has to be first.

This forces two placement rules:

• The signal peptide occupies the N-terminus, so an N-terminal affinity tag either has to sit between the signal and the mature protein (and it will be exposed at the new N-terminus after signal cleavage, which may or may not be what you want) or it conflicts with the signal entirely.
• For secreted constructs, put the affinity tag at the C-terminus. The signal is cleaved off during export, so any N-terminal tag placed upstream of the mature protein is either removed with the signal or ends up at the processed N-terminus in a way that is hard to control. A C-terminal tag survives translocation intact and reports on fully secreted, mature protein. (The secreted-glycoprotein case in the affinity-tag post is exactly this failure and fix.)

Membrane Protein Topology: Which Terminus Is Cytoplasmic?

For membrane proteins, the terminus question becomes a topology question: is this end inside the cell (cytoplasmic) or outside (periplasmic or extracellular)? The answer determines whether a tag can even be captured, whether it folds, and in the worst case whether it flips the protein's orientation in the membrane.

Topology follows the positive-inside rule: loops rich in Arg and Lys stay on the cytoplasmic side, and this charge bias is the strongest single determinant of how a polytopic membrane protein threads the bilayer (von Heijne, 1992). A tag adds sequence, and a charged tag near a marginal topogenic signal can tip a borderline segment, changing the topology you were trying to study.

Placement also interacts with tag folding. In the proteome-wide topology mapping of Daley et al. (2005), C-terminal GFP and PhoA fusions were used as topology reporters precisely because GFP folds and fluoresces only in the cytoplasm while PhoA is active only in the periplasm. The same physics constrains your affinity tag: a tag that requires cytoplasmic folding (or a disulfide-dependent tag that needs an oxidizing periplasm) will behave differently depending on which compartment its terminus lands in. Before tagging a membrane protein, you need the predicted topology, then you place the tag on the terminus whose compartment is compatible with both capture and the tag's own folding requirements.

The C-Terminal Motifs a Tag Will Quietly Destroy

Here is where C-terminal tagging goes from "suboptimal" to "silently wrong." A large fraction of eukaryotic targeting and interaction signals are C-terminal, and they require a free, unmodified C-terminus. Bolt a tag on and you delete the signal without any obvious sign at the bench, until the localization, interaction, or activity you were counting on is gone.

The recurring offenders:

• CAAX prenylation. Ras-family GTPases and many other proteins end in a CAAX box that is prenylated (farnesyl or geranylgeranyl), then trimmed and carboxymethylated, to drive membrane association (Zhang & Casey, 1996). The modification happens on the C-terminal cysteine. A C-terminal tag makes the CAAX box internal and it is never prenylated, so your GTPase never reaches the membrane.
• PDZ-binding motifs. PDZ domains recognize the last few residues of a partner protein and, critically, the free C-terminal carboxylate (Songyang et al., 1997). Add a C-terminal tag and the interaction is abolished. If you are studying a receptor or channel that docks to a PDZ scaffold, tag the N-terminus.
• KDEL / HDEL ER retention. Soluble ER-resident proteins carry a C-terminal KDEL that the KDEL receptor reads to retrieve them from the Golgi (Munro & Pelham, 1987). A C-terminal tag buries KDEL and the protein is secreted instead of retained.
• Tail-anchored membrane proteins. A whole class of proteins (SNAREs, many Bcl-2 members, cytochrome b5) insert into membranes via a single C-terminal transmembrane segment inserted post-translationally (Borgese & Fasana, 2011). The targeting information is the C-terminal tail itself. A C-terminal tag blocks insertion entirely.

The rule that falls out: if your protein's function or localization lives at the C-terminus, tag the N-terminus, and vice versa. The corollary is that you have to know whether a functional motif sits at a terminus before you design, which is a characterization step, not a cloning step.

Purification Accessibility and Full-Length Selection

Two placement effects are purely about getting clean protein out.

Accessibility. A tag can only be captured if it is exposed. Termini are usually solvent-accessible (Jacob & Unger, 2007), but a terminus tucked into the fold, buried at an oligomer interface, or occluded by an adjacent domain will bind resin poorly no matter how good the tag chemistry is. If the predicted structure shows one terminus buried, tag the other.

Full-length selection. Because translation proceeds N to C, a C-terminal tag is present only on molecules that were translated all the way through. Capturing by a C-terminal tag therefore selects against prematurely terminated and internally clipped species: truncation products lack the tag and flow through. An N-terminal tag captures slightly more (the tag is made first and usually exposed) but co-purifies any C-terminally truncated fragments. For a protein prone to internal stops, ribosome drop-off, or C-terminal proteolysis, C-terminal capture is a free purity win.

These two effects can pull in opposite directions (accessibility might favor N, full-length selection favors C), which is why placement is a judgment call informed by the specific protein, not a blanket rule.

Cleavage and Scar Residues Depend on the Terminus

If you plan to remove the tag, the terminus determines what scar you are left with, because proteases cut with a defined polarity.

Most site-specific proteases cleave leaving their downstream residues attached to whatever is C-terminal of the cut. For an N-terminal tag, the construct reads tag - protease site - protein, so cleavage leaves the protease's scar residues on the new N-terminus of your protein (for example, TEV's ENLYFQ|S leaves a serine; HRV 3C's LEVLFQ|GP leaves Gly-Pro). For a C-terminal tag, the construct reads protein - protease site - tag, and the arithmetic flips: the recognition sequence itself is left on your protein's C-terminus while the tag departs with the downstream residues.

Consequences:

• If your application needs a native N-terminus (many structural and functional studies), an N-terminal tag plus a scar-free strategy (SUMO protease, which cuts by recognizing the SUMO fold and leaves no scar) is the clean route. We go deep on scar residues and protease choice in the tag-removal guide.
• If you need a native C-terminus (for example, to preserve a C-terminal function that you nonetheless had to keep tagged during purification), you have to think carefully, because standard C-terminal protease geometry tends to leave more of the recognition sequence behind.
• The N-end rule (above) means the scar residue you expose at an N-terminus is not just cosmetic, it can set the protein's half-life.

Tandem and Dual Tags: Use Both Ends on Purpose

When one tag is not clean enough, the most powerful move is to put two orthogonal tags on opposite termini: for example, His at the N-terminus and Strep at the C-terminus. Purifying by each in sequence does two things. It removes contaminants that sneak through the first capture (they are chemically unrelated to the second chemistry), and because the two tags sit at opposite ends, the sequence of captures selects for only doubly-tagged, full-length molecules: anything truncated at either end loses one tag and drops out. This is the workhorse dual-tag strategy for structural biology and is discussed as a tag decision in the affinity-tag post; here the point is that it is fundamentally a placement strategy, one tag per terminus, chosen deliberately.

The cost is footprint (two tags is more sequence to perturb folding and more to remove) and yield (each step loses material). Split-terminus dual tags are for when purity is non-negotiable, not a default.

Case Study: The GTPase That Never Reached the Membrane

Problem. A team studying a small Ras-family GTPase needed clean, functional protein for a membrane-recruitment assay. Following lab habit, they used a vector with a C-terminal His10 tag. Expression and IMAC purification looked fine: good yield, clean gel. But in the assay the protein stayed soluble and never associated with the membrane fraction, and the results made no sense.

Analysis. The protein terminated in a CAAX box. C-terminal prenylation of that terminal cysteine is what drives membrane association (Zhang & Casey, 1996). By tagging the C-terminus, the team had made the CAAX box internal, so it was never prenylated and the protein had no membrane anchor. The purification "worked" precisely because the protein stayed soluble, which was the bug, not a feature. Nothing on the gel could have revealed this, because the tag did not affect expression or folding, only the buried C-terminal signal.

Solution. Move the tag to the N-terminus, leaving the C-terminal CAAX box free and unmodified, and express in a system competent for prenylation. The N-terminus of this GTPase carried no functional motif and was solvent-exposed in the predicted structure, so it tolerated the tag without perturbing the fold or the nucleotide-binding site.

Outcome. The N-terminally tagged construct was prenylated, associated with the membrane fraction, and behaved correctly in the recruitment assay. The fix was a terminus swap decided at the design stage. Had the C-terminal motif been flagged before cloning, the team would have saved a month and a confusing dataset.

Practical Tool: A Complete Decision Tree for Choosing the Terminus

Work top to bottom. The first rule that applies wins, because the hard constraints (secretion, C-terminal motifs, topology) override the soft preferences (expression, purity).

START: Choose the terminus for your tag.

STEP 1: Is the protein secreted or exported (has an N-terminal signal peptide)?
   YES → Signal peptide occupies the N-terminus (mandatory).
         → Put the AFFINITY TAG at the C-TERMINUS. Done.
   NO  → continue.

STEP 2: Does the protein have a functional/targeting motif at a terminus?
   (CAAX box, PDZ-binding tail, KDEL/HDEL, tail-anchor TM segment = C-terminal;
    rare N-terminal targeting or processing signals = N-terminal)
   C-terminal motif present → tag the N-TERMINUS. Done.
   N-terminal motif present → tag the C-TERMINUS. Done.
   None → continue.

STEP 3: Is it a membrane protein?
   YES → Get the predicted topology. Tag the terminus whose compartment
         (cytoplasmic vs periplasmic/extracellular) is compatible with BOTH
         capture and the tag's own folding needs. A charged tag near a
         marginal topogenic signal can flip topology: prefer the terminus
         that does not sit next to a borderline TM segment. Then continue.
   NO  → continue.

STEP 4: Do you need a SOLUBILITY FUSION (protein is insoluble)?
   YES → Solubility fusions must fold first: place them at the N-TERMINUS.
         (This effectively chooses the terminus.) Done.
   NO  → continue.

STEP 5: Is one terminus buried or at a domain/oligomer interface in the
         predicted structure?
   YES → Tag the OTHER (exposed) terminus. Continue if both are exposed.

STEP 6: Is the terminus near the active site or a binding surface?
   YES → Avoid that terminus; tag the functionally silent one.

STEP 7: Now optimize the soft goals on whichever termini remain:
   • Fighting for EXPRESSION level in E. coli?  Prefer C-TERMINAL
     (keeps the 5' mRNA / translation-initiation region clean), and
     codon-optimize the tag+linker region regardless.
   • Prone to TRUNCATION / need FULL-LENGTH only?  Prefer C-TERMINAL
     (captures only fully translated molecules).
   • Need maximum CAPTURE efficiency on a difficult/dilute target?
     N-TERMINAL captures slightly more (fragments included).
   • Need a NATIVE N-TERMINUS after cleavage?  N-terminal tag + a
     scar-free removal (SUMO fold-recognition protease).
   • Purity NON-NEGOTIABLE?  Split a dual tag across BOTH termini
     (e.g., His at N + Strep at C) to select full-length, doubly-tagged.

STEP 8: Verify the assembled sequence: check the processed N-terminus
   (initiator-Met excision + N-end-rule residue after any cleavage),
   the scar residues at the tagged terminus, and the codon-optimized DNA.

The Economics of Getting the Terminus Wrong

Terminus placement costs nothing to decide correctly and a month to decide wrong.

Getting it wrong (the common path):

Getting it right (design first):

The difference is five weeks and a dataset you can trust. The expensive part was never the reagents; it was the failed campaign and the misleading "clean gel" that hid a deleted C-terminal signal. Make the decision where it is cheap to change your mind: in the construct, not at the column.

Bottom Line

The terminus you tag is a biological decision, not a formatting one. Hard constraints come first and override everything else: signal peptides force the tag to the C-terminus; C-terminal functional motifs (CAAX, PDZ-binding, KDEL, tail-anchors) force it to the N-terminus; membrane topology dictates which end is even accessible and foldable. Only after the constraints are satisfied do you optimize the soft goals: C-terminal placement protects the translation-initiation region and enriches for full-length protein, N-terminal placement is mandatory for solubility fusions and captures slightly more, and a dual tag split across both ends buys full-length purity when it is non-negotiable. The one thing you cannot do is default to whichever terminus your vector happens to offer and hope the protein does not mind. Sometimes it minds a great deal, and it will not tell you on the gel.

How Orbion Helps You Choose the Terminus

Terminus placement is a design decision, and it depends on things you cannot see in a sequence alone: where the functional motifs are, which terminus is buried, and how a membrane protein is oriented. Orbion turns that from a guess into a check.

Start in Characterization. AstraUNFOLD predicts transmembrane topology (so you know which terminus is cytoplasmic vs periplasmic), plus per-residue disorder and amyloidogenicity to tell you whether a terminus is exposed and safe to load. AstraBIND flags ligand-binding sites so you do not tag next to the active site. AstraPTM predicts modification sites (including terminus-proximal ones) so you can keep a tag away from residues that must stay modified. Together these tell you which terminus is safe to tag before you design anything.

Then build in the Design tab. Construct both the N-terminal and the C-terminal version from the component library: pick from 17 affinity tags with N- or C-terminal placement and cleavage options, add a signal peptide (N-terminal, from the 14 in the library) for secreted constructs, attach a solubility or stabilization fusion, and set linkers. Orbion scores each construct on solubility, disorder, aggregation, ΔTm (via AstraDTM) and ΔΔG (via AstraDDG), and reports PTM-site retention, so you can see side by side whether the N- or C-terminal placement perturbs the fold or the modifications. It matches each construct against your organization's vector library, and hands the winner off to Bench for construct-aware protocols.

Finally, check the assembly. Orbion shows the exact assembled protein sequence for each terminus choice, including the scar residues left by your chosen protease, alongside the codon-optimized DNA (with 5' structure and Shine-Dalgarno-like sequences handled for the target host). You see both the N- and C-terminal versions in full before you order a single oligo.

Decide the terminus where it is cheap to change your mind. Upload your sequence to Orbion and compare both ends before you clone.

References

1. Bachmair, A., Finley, D., & Varshavsky, A. (1986). In vivo half-life of a protein is a function of its amino-terminal residue. Science, 234(4773), 179-186. https://doi.org/10.1126/science.3018930

2. Varshavsky, A. (2011). The N-end rule pathway and regulation by proteolysis. Protein Science, 20(8), 1298-1345. https://doi.org/10.1002/pro.666

3. Kudla, G., Murray, A.W., Tollervey, D., & Plotkin, J.B. (2009). Coding-sequence determinants of gene expression in Escherichia coli. Science, 324(5924), 255-258. https://doi.org/10.1126/science.1170160

4. Goodman, D.B., Church, G.M., & Kosuri, S. (2013). Causes and effects of N-terminal codon bias in bacterial genes. Science, 342(6157), 475-479. https://doi.org/10.1126/science.1241934

5. Frottin, F., Martinez, A., Peynot, P., Mitra, S., Holz, R.C., Giglione, C., & Meinnel, T. (2006). The proteomics of N-terminal methionine cleavage. Molecular & Cellular Proteomics, 5(12), 2336-2349. https://doi.org/10.1074/mcp.M600225-MCP200

6. Woestenenk, E.A., Hammarström, M., van den Berg, S., Härd, T., & Berglund, H. (2004). His tag effect on solubility of human proteins produced in Escherichia coli: a comparison between four expression vectors. Journal of Structural and Functional Genomics, 5(3), 217-229. https://doi.org/10.1023/B:JSFG.0000031965.37625.0e

7. von Heijne, G. (1992). Membrane protein structure prediction. Hydrophobicity analysis and the positive-inside rule. Journal of Molecular Biology, 225(2), 487-494. https://doi.org/10.1016/0022-2836(92)90934-C

8. Daley, D.O., Rapp, M., Granseth, E., Melén, K., Drew, D., & von Heijne, G. (2005). Global topology analysis of the Escherichia coli inner membrane proteome. Science, 308(5726), 1321-1323. https://doi.org/10.1126/science.1109730

9. Munro, S., & Pelham, H.R.B. (1987). A C-terminal signal prevents secretion of luminal ER proteins. Cell, 48(5), 899-907. https://doi.org/10.1016/0092-8674(87)90086-9

10. Zhang, F.L., & Casey, P.J. (1996). Protein prenylation: molecular mechanisms and functional consequences. Annual Review of Biochemistry, 65, 241-269. https://doi.org/10.1146/annurev.bi.65.070196.001325

11. Songyang, Z., Fanning, A.S., Fu, C., Xu, J., Marfatia, S.M., Chishti, A.H., Crompton, A., Chan, A.C., Anderson, J.M., & Cantley, L.C. (1997). Recognition of unique carboxyl-terminal motifs by distinct PDZ domains. Science, 275(5296), 73-77. https://doi.org/10.1126/science.275.5296.73

12. Jacob, E., & Unger, R. (2007). A tale of two tails: why are terminal residues of proteins exposed? Bioinformatics, 23(2), e225-e230. https://doi.org/10.1093/bioinformatics/btl318

13. Borgese, N., & Fasana, E. (2011). Targeting pathways of C-tail-anchored proteins. Biochimica et Biophysica Acta (Biomembranes), 1808(3), 937-946. https://doi.org/10.1016/j.bbamem.2010.07.010